Dose-response relationship of lung cancer to amount smoked, duration and age starting

Table 4 Amount smoked by current smokers - observed and fitted lung cancers for the linear, best power and best log-with-baseline model, with β₁ fitted separately for each block

Midpoint amount smoked (cigs/d)	Observed1	Fitted2
		Linear model	Best power model	Best log-with-baseline model
< 5	1249.17	1023.00	1297.42	1173.88
5 to < 10	2579.62	2156.31	2595.37	2539.78
10 to < 15	6125.69	4749.92	6276.74	6299.68
15 to < 20	7940.18	6678.14	8009.06	8156.50
20 to < 30	18138.36	15724.45	17468.99	17678.51
30 to < 40	3858.94	4106.12	3743.23	3701.31
40+	7703.88	8860.42	8115.77	7949.95
Never smoked	6649.61	10947.08	6738.86	6745.84
Total	54245.45	54245.45	54245.45	54245.45
Fit statistic3		3792.53	65.29	56.78

¹Observed pseudo-number of lung cancer cases, summed over blocks;

²Fitted pseudo-number of lung cancer cases, summed over blocks. For each study, the fitted number of cases for each block is calculated from the numbers at risk and the fitted RRs by amount smoked, derived from the fitted value of β₁;

³Based on summation of (observed-fitted)²/fitted, the summation also including terms for the observed and fitted total numbers of controls (not shown). The statistic can be considered to be approximately chisquared on 12 DF and is significant at P < 0.001 for all three models.